config.json

{
    "_name_or_path": "facebook/dinov2-large",
    "apply_layernorm": true,
    "architectures": [
        "Dinov2ForImageClassification"
    ],
    "attention_probs_dropout_prob": 0.0,
    "drop_path_rate": 0.0,
    "hidden_act": "gelu",
    "hidden_dropout_prob": 0.0,
    "hidden_size": 1024,
    "id2label": {
        "0": "Acropore_branched",
        "1": "Acropore_digitised",
        "2": "Acropore_tabular",
        "3": "Algae",
        "4": "Dead_coral",
        "5": "Millepore",
        "6": "No_acropore_encrusting",
        "7": "No_acropore_massive",
        "8": "No_acropore_sub_massive",
        "9": "Rock",
        "10": "Rubble",
        "11": "Sand"
    },
    "image_size": 140,
    "initializer_range": 0.02,
    "label2id": {
        "Acropore_branched": 0,
        "Acropore_digitised": 1,
        "Acropore_tabular": 2,
        "Algae": 3,
        "Dead_coral": 4,
        "Millepore": 5,
        "No_acropore_encrusting": 6,
        "No_acropore_massive": 7,
        "No_acropore_sub_massive": 8,
        "Rock": 9,
        "Rubble": 10,
        "Sand": 11
    },
    "layer_norm_eps": 1e-06,
    "layerscale_value": 1.0,
    "mlp_ratio": 4,
    "model_type": "dinov2",
    "num_attention_heads": 16,
    "num_channels": 3,
    "num_hidden_layers": 24,
    "out_features": [
        "stage24"
    ],
    "out_indices": [
        24
    ],
    "patch_size": 14,
    "problem_type": "multi_label_classification",
    "qkv_bias": true,
    "reshape_hidden_states": true,
    "stage_names": [
        "stem",
        "stage1",
        "stage2",
        "stage3",
        "stage4",
        "stage5",
        "stage6",
        "stage7",
        "stage8",
        "stage9",
        "stage10",
        "stage11",
        "stage12",
        "stage13",
        "stage14",
        "stage15",
        "stage16",
        "stage17",
        "stage18",
        "stage19",
        "stage20",
        "stage21",
        "stage22",
        "stage23",
        "stage24"
    ],
    "torch_dtype": "float32",
    "transformers_version": "4.41.0",
    "use_swiglu_ffn": false,
    "initial_learning_rate": 0.001,
    "train_batch_size": 64,
    "eval_batch_size": 64,
    "optimizer": {
        "type": "Adam"
    },
    "lr_scheduler_type": {
        "type": "ReduceLROnPlateau"
    },
    "patience_lr_scheduler": 5,
    "factor_lr_scheduler": 0.1,
    "weight_decay": 0.0001,
    "early_stopping_patience": 10,
    "freeze_encoder": true,
    "data_augmentation": true,
    "num_epochs": 150
}